Tracing the Origins of Algorithmic Life
Roots of ‘big data’ can be seen in the history of Japanese psychological tests
You’re being tested. You don’t know the criteria used to determine your score—or even your results. The test is being administered not by a human teacher or moderator, but by machines. And it’s going on 24 hours a day, every day of your life.
“The use of big data and algorithms in machine learning is a kind of test,” says 2024 Harvard Horizons Scholar Juhee Kang. “It’s going on constantly and more than ever we see new forms of artificial intelligence used in calculating and measuring different types of aptitudes.”
A graduating PhD student in history and East Asian languages and civilizations at the Harvard Kenneth C. Griffin Graduate School of Arts and Sciences (GSAS), Kang explores how testing and mass data collection evolved in 20th-century Japan, where they became central across society. In her Horizons project, “Numbers, Minds, and Society,” Kang charts the path of psychological tests from scientific novelty to the bulwark of “scientific management” and meritocracy. Her findings lead her neither to condemn nor promote testing but to call for transparency and public discussion as humanity enters an age governed by algorithms.
“Tesuto”
Lee and Juliet Folger Fund Professor of History Andrew Gordon, one of Kang’s faculty advisors, says psychological tests have become deeply interwoven into societies and cultures throughout the world. They remain both widely used and controversial, for instance, as a factor in determining admissions to prestigious universities. “Understanding when and how psychological tests emerged globally is both a fascinating historical question and one with implications for the present,” he says. Moreover, Japan is a particularly interesting subject for exploration because, unlike China, it did not have a long premodern history of using examinations to evaluate candidates for high office, and because of its international engagement.
“In the 20th century, as Juhee shows, the nation—including corporations, government offices, and commercial test creators—broadened the understanding of the role and value of tests of many sorts,” Gordon says. “They were seeking the holy grail of a test that did not examine subject knowledge but deeper attributes of individuals. This took place through engagement with psychologists and test advocates in the West, especially Germany and the US, so this is a global story.”
Kang became curious about the history of testing in Japan, the eastern neighbor of her native Korea and the first non-Western industrial and imperial power. She says that beginning in the early 20th century, Japanese companies and institutions used psychological tests to determine who qualified for jobs and educational opportunities.
In the 20th century . . . [the Japanese] broadened the understanding of the role and value of tests of many sorts. They were seeking the holy grail of a test that did not examine subject knowledge but deeper attributes of individuals.
—Professor Andrew Gordon
“In the 1890s, you find the first use of the term ‘tesuto’ in a Japanese scientific journal,” she says. “I realized they were referring to something fundamentally different from the mere act of evaluating people based on some criteria, or even any historical precedents of official selection such as the civil service examination, and which only a very small portion of the population was even allowed to take. They were referring to a kind of scientific experiment that involved building a system of data collection, analysis, and formulation of a standard based on the information gathered.”
By the 1920s, Japan, like a lot of industrialized nations, was experiencing a surge in social movements. One such movement was the drive for efficiency in business operations. “‘Scientific’ management was a big force globally in the early 20th century,” Kang says. “One of its central principles was testing people based on their aptitude to perform a task and then putting them where they could be maximally productive.”
At the same time, the new education movement in Japan emphasized the innate qualities of individuals. The idea was to foreground a person’s distinctive characteristics so they could fully manifest their natural talents and growth.
“There comes from the education movement a criticism that’s very familiar to us today,” Kang says. “Namely, that tests are biased toward those with a particular cultural knowledge. If you are given the privilege of being born into a certain family or being part of a social group, you learn, with only moderate effort, the qualities and skills the privileged class highly values.”
Scientists in the first decades of the 20th century sought to resolve the problem by arguing for the use of psychological tests, for which, they argued, a subject could not prepare. It was around this time that knowledge about psychological tests—including the IQ test—had been imported for Japanese scientists to study and modify for use in Japan.
“Psychologists claimed that some of these psychological tests could measure something truly innate that was unaffected by all kinds of environmental factors,” Kang says. “Educators were divided. One faction welcomed the move; another resisted, saying that any form of test would cause the same problem in the end. Still, another consisted of those who believed in the efficacy of psychological tests but worried that it would make personal effort worthless.”
Psychological tests emerged as a new cutting-edge technology of human differences precisely when Japan’s existing social structure could no longer accommodate its growing diversity.
—Juhee Kang
The Contour of a Person’s Mind
One of the measures Kang explores in depth is the Uchida-Kraepelin (UK) test. Inspired by the work of German psychiatrist Emil Kraepelin, Japanese psychologist Uchida Yuzaburo in 1926 developed a new way to represent the pattern of a person’s learning process. Subjects got a piece of paper that had 25 horizontal rows. Each row contained a series of single-digit numbers, one after the other. Subjects added the numbers in each row for one minute, writing down only the last digit of each sum on the paper as they proceeded. For instance, say the first two numbers in a row were 6 and 7. The sum would be 13, so the subject would write “3” in the space next to the 7 and then do the same for the next two numbers in the row. The aim was to do as many sums as possible in a minute, take a short break, and then do it again for the next row until the subject finished the first 15 rows. Then they got a 5-minute break before starting again on the final 10 rows.
“The scientists wanted a simple version of the advanced mental processes that we are required to do unconsciously to live our daily lives,” Kang explains. “Then, based on repetition of that simple form of work, the test would measure something fundamental, something that's so unique about this person's psyche that they could argue that it was the contour of a person's mind.”
In this case, Kang uses the word “contour” quite literally. When a subject had completed the test, scientists drew a line that connected the last sums in each of the 25 rows: a work performance curve. “At the beginning of the test, participants were usually excited about the new activity, so the curve went up,” Kang says. “Then, because just adding is such tedious work, performance declined. As subjects approached the end, they got excited again and performance rose once more. Based on those assumptions, scientists judged the ‘normal’ and ‘abnormal’ pattern of this curve and applied it to their assessments.”
The UK test was increasingly used to measure aptitude in a variety of settings as its commercial uses became evident. "In the 1920s, the test was developed as an alternative to scholastic exams,” Kang says. “In the 1930s, however, it became more widely applied to occupational selections, mainly for 'modern’ jobs that required lots of repetition, such as telegraphers, train or taxi drivers, and telephone operators. In the 1940s, it then took a darker turn and was used to screen malingerers among the repatriated soldiers who worked at munitions factories to address wartime labor shortages. Sometime in the 1970s, it was exported as the human resources management tool that guaranteed ‘Japan quality’ at manufacturing factories in Southeast Asia."
Kang says the UK test matters because of its claimed ability to relieve the modern anxiety of not intuitively knowing oneself and others. “As Japan became a multiracial, multiethnic powerhouse, it also became a highly industrialized society, in which people from diverse backgrounds began to live side by side,” she says. “As physical differences became more evident and ‘knowing’ a person got more complex, people sought technology that could manifest the invisible qualities. Scientific tests assured them with their purported methodological objectivity that aimed to transcend linguistic, cultural, or other environmentally induced biases in assessment.”
Discretionary Effort
Kang is not an opponent of psychological testing. She says humans living in the whirl of modernity need some kind of discretionary system. “Modern society is inevitably based on some form of judgment and discrimination,” Kang says. “Testing has come to serve that function. In fact, psychological tests emerged as a new cutting-edge technology of human differences precisely when Japan’s existing social structure could no longer accommodate its growing diversity.”
At the same time, Kang is alarmed by the constancy of algorithmic testing and the way that human life is increasingly shaped by artificial intelligence (AI). She sees parallels between early 20th-century Japanese society and today. Both are worlds of increasing visible diversity that give rise to new assessment technologies.\
“We are the subject of algorithmic testing by big tech companies whether we like it or not,” she says. “To use their services, we have to click a little box that basically says, ‘I agree to allow you to do whatever you want with my data, and I don’t need to know anything about it.’ But, increasingly, there are more concerns about what’s going on inside the ‘black box’ of how AI learns and creates.”
Kang says that as with any science or technology, the benevolence or malignance of algorithmic testing and machine learning is a function of the value systems underpinning them. That’s why she believes that transparency is crucial in the formulation of these technologies—as is a thoroughgoing public conversation about their implementation.
“Digital life needs to be in balance with the physical world—actual human interactions,” she says. “Some things are untouchable by testing. Some things are so uniquely personal, so uniquely valuable that statistically based judgment has no meaning in them.”
Get the Latest Updates
Subscribe to Colloquy Podcast
Simplecast Stitcher
